{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/deepskeleton-learning-multi-task-scale","title":"DeepSkeleton: Learning Multi-task Scale-associated Deep Side Outputs for Object Skeleton Extraction in Natural Images","arxiv_id":"1609.03659","date":"2016-09-13","proceeding":null,"authors":["Wei Shen","Kai Zhao","Yuan Jiang","Yan Wang","Xiang Bai","Alan Yuille"],"abstract":"Object skeletons are useful for object representation and object detection.\nThey are complementary to the object contour, and provide extra information,\nsuch as how object scale (thickness) varies among object parts. But object\nskeleton extraction from natural images is very challenging, because it\nrequires the extractor to be able to capture both local and non-local image\ncontext in order to determine the scale of each skeleton pixel. In this paper,\nwe present a novel fully convolutional network with multiple scale-associated\nside outputs to address this problem. By observing the relationship between the\nreceptive field sizes of the different layers in the network and the skeleton\nscales they can capture, we introduce two scale-associated side outputs to each\nstage of the network. The network is trained by multi-task learning, where one\ntask is skeleton localization to classify whether a pixel is a skeleton pixel\nor not, and the other is skeleton scale prediction to regress the scale of each\nskeleton pixel. Supervision is imposed at different stages by guiding the\nscale-associated side outputs toward the groundtruth skeletons at the\nappropriate scales. The responses of the multiple scale-associated side outputs\nare then fused in a scale-specific way to detect skeleton pixels using multiple\nscales effectively. Our method achieves promising results on two skeleton\nextraction datasets, and significantly outperforms other competitors.\nAdditionally, the usefulness of the obtained skeletons and scales (thickness)\nare verified on two object detection applications: Foreground object\nsegmentation and object proposal detection.","url_abs":"http://arxiv.org/abs/1609.03659v3","url_pdf":"http://arxiv.org/pdf/1609.03659v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"deepskeleton-learning-multi-task-scale","repo_url":"https://github.com/zeakey/skeleton","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"multi-task-learning","task_name":"Multi-Task Learning"},{"task_slug":"object","task_name":"Object"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[],"datasets_introduced":[{"slug":"sk-large","name":"SK-LARGE","full_name":"SK-LARGE"}],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1609.03659","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}